Simulation of Hypervelocity Scramjet Combustion with

Oxygen Enrichment

D. Petty 1, M. K. Smart 1, V. Wheatley 1 and S. A. Razzaqi 1

1 Centre for Hypersonics, The University of Queensland, Brisbane, Queensland, 4072,

Australia

Summary: A numerical investigation of oxygen enriched combustion within a hypervelocity

scramjet was performed using two-dimensional Reynolds-averaged Navier-Stokes

simulations. The simulations modelled previous oxygen enrichment experiments in a

simplified scramjet flow-path. It is shown that oxygen enrichment significantly affects the

mixing and combustion characteristics within a scramjet combustor. Simulation results

indicate that combustion efficiency of the scramjet improved beyond the amount expected

from the O2 premixed with the fuel reacting with the stoichiometric quantity of H2.

Keywords: hypervelocity, scramjet, oxygen enrichment, air-breathing.

Introduction

Air-breathing propulsion systems, such as the scramjet engine, with access-to-space capability

have attracted considerable interest due to the current dependency on multi-stage rockets with

low specific impulse. A three-stage rocket-scramjet-rocket system has been proposed to power

a launch vehicle into a low Earth orbit [1]. The launch vehicle will travel at near-constant

dynamic pressure during the operation of the scramjet. Results from the study showed that

such a vehicle would struggle to maintain net thrust above Mach 12. This restricts the

usefulness of scramjet engines for space access.

The difficulty of scramjet engines maintaining net thrust at higher altitudes is contributed by

both air (specifically O2) mass capture and residence time of O2 within the combustor

decreasing. These effects can be illustrated by performing a simple analysis. Air density of the

atmosphere decreases by three orders of magnitude when travelling from the Earth’s surface to

an altitude of 47km [2]. Species mass fractions remain effectively constant within these

altitudes so the mass of O2 per unit volume of air decreases linearly with air density. Dynamic

pressure variations are minimal in order to generate sufficient lift whilst avoiding excessive

drag and heating loads [3]. Using the definition of dynamic pressure (q) and the one-

dimensional continuity equation, the mass flow capture (ṁC) of a scramjet engine can be

defined as:

2 CC

q Am

v

(1)

This relationship shows that for trajectories which traverse along near constant dynamic

pressure paths, the mass flow capture of the engine is inversely related to the flight speed (v∞).

Variations in capture area (AC) are trivial in comparison to the increase in flight speed as the

vehicle accelerates so mass capture must decrease with increasing altitude. The adverse effect

of diminishing O2 mass capture on scramjet performance is further compounded by the

decreasing residence time of flow within the scramjet engine as flight speeds increase. It can

be concluded from these factors that achieving high combustion efficiency is paramount in

hypervelocity scramjet engines. It is critical that effective mixing of air and fuel at the

molecular level to a near stoichiometric mixture of O2 and fuel occurs [3].

Oxygen enrichment is a technique aimed at supplementing the diminishing amount of O2

capture at altitudes and speeds where scramjet shutoff is predicted to occur. The addition of

O2 has two additional effects at any altitude: the injected flow shifts towards a stoichiometric

mixture of fuel and oxidizer as well as changing the fluid properties of the injected flow [4].

Pike [5] showed for a restricted class of scramjets that the addition of O2 to the injected fuel

decreased launch mass of a vehicle whilst increasing the payload mass.

Shock tunnel experiments simulating hypervelocity conditions were conducted by Razzaqi &

Smart [6] to investigate the effectiveness of oxygen enrichment within a simplified scramjet

engine. Oxygen enrichment was found to be effective at increasing the thrust of the scramjet,

particularly at high altitudes. However, as only pressure transducer point measurements were

taken, the details of the flow physics that led to the improved performance were not presented.

This paper attempts to reveal the details of the flow physics by carrying out a numerical

investigation of Razzaqi and Smart’s experiments. This report is structured as follows: first we

present an overview of the Razzaqi and Smart experiment [6], then the numerical

methodology employed for this study is described and finally, the results obtained are

presented, focusing on combustion efficiency comparisons.

Oxygen Enrichment Experiments

The Razzaqi & Smart experiments were performed in the T4 free piston reflected shock tunnel

at The University of Queensland [7]. The experiment utilized a contoured axisymmetric

nozzle to produce two effective flight altitude conditions: standard altitude with M∞ = 12.3

and q∞ = 40.2 kPa and high altitude with M∞ = 11.4 and q∞ = 9.8 kPa. The numerical

simulations performed for this study focused specifically on the standard altitude conditions.

Two high range PCB pressure transducers were mounted in the stagnation region to record the

nozzle-supply pressure [6]. A summary of the selected experimental shock tunnel conditions

for simulation is detailed in Table 1.

Table 1: Summary of selected shock tunnel test conditions

Shot Inflow Composition Temp

(K)

Pres.

(kPa)

Mach No.

(-/-)

Mass Flow

(kg/sec)

9941 Nozzle Exit Air 988 6.3 5.7 N/A

Injector Off N/A N/A N/A N/A

9952 Nozzle Exit N2 958 5.9 5.9 N/A

Injector H2 300 (t) 300 (t) ~ 1 0.0193

9953 Nozzle Exit Air 986 6.4 5.7 N/A

Injector H2 (ϕ=0.84) 300 (t) 301 (t) ~ 1 0.0194

9944 Nozzle Exit Air 950 6.11 5.7 N/A

Injector H2 (ϕ=0.81)

+ Oxy (EP~15%)

300 (t) 471 (t) ~ 1 0.0183

+ 0.0272

Sequentially, these shots correspond to air only (no fuel injection), mixing only (fuel injected

into N2), combustion, and combustion with oxygen enrichment. The fuelling conditions of

both simulation and experiment are described using the standard definition of fuel equivalence

ratio (ϕ) for hydrogen fuel:

2 2

2

,1

2 0.23

O H IN

H C

M m

M m (2)

Where 2OM is the molar mass of O2,

2HM is the molar mass of H2, 2 ,H INm is the mass flow

rate of injected H2, Cm is the mass flow rate of ingested O2.

The oxygen enrichment shot employed an additional flow (Oxy) of 75/25 (% by wt) O2/N2

mixture which was premixed with the fuel flow before injection. The parameter used to

describe the amount oxygen enrichment is equivalent to the Enrichment Percentage (EP) used

by Razzaqi & Smart [6]. This parameter is defined as the percentage of injected H2 which

would be consumed in a stoichiometric reaction with the injected enrichment oxygen:

2 2

2 2

,

,

2 100%H O IN

O H IN

M mEP

M m (3)

Where 2 ,O INm is the mass flow rate of injected O2. The objective of this set of experiments was

to investigate the effect of varying EP whilst holding 2 ,H INm and Cm relatively constant.

The experimental model consists of three primary features: a two-dimensional compression

wedge at an angle of 8° to the flow, a constant area rectangular combustor (duct) that has an

aspect ratio of l/h = 18.4 and an expansion surface inclined at an angle 11° to the flow

direction

The duct has a height of 47 mm and a width of 100 mm. An injector strut with a height of

7 mm injects fuel through a 1 mm × 100 mm slot located on the backward face. A cross-

sectional schematic of the experimental test section is shown in the Fig. 1.

Fig. 1: Experimental model schematic [6]

The findings of the experimental campaign of [6] validated the previous theoretical findings,

showing that specific thrust linearly increases with equivalence ratio at a particular oxygen

enrichment percentage.

Determination of combustion efficiency from the experimental data required the use of a quasi

one-dimensional cycle analysis method [8]. The analysis assumes a thermally perfect mixture

of gases in thermal equilibrium. Skin friction drag and heat transfer across the duct walls were

approximated using the Reynolds analogy. A combustion efficiency curve, taken from Heiser

and Pratt [3], was used to model the effective mixing and kinetics of combustion

Employing this cycle analysis to predict pressure values along the duct, the combustion

efficiency curve was manipulated to best match the experimental pressure measurements taken

along the lower surface of the duct (Kulite transducers). Once pressure values were matched,

the derived combustion efficiency curve was considered representative of the true combustion

within the experimental duct, as shown in Fig. 2.

Fig. 2: Typical cycle predicted and experimental pressure distributions [8]

Using this method, the combustion efficiencies for shots 9953 and 9944 were found to be 45%

and 60% respectively. It was concluded that combustion efficiency increased by an amount

equivalent to the enrichment percentage.

Numerical Methodology

NASA’s viscous upwind algorithm for complex flow analysis (VULCAN) version 6.1.0 [9]

was the numerical solver used in this investigation. Vulcan can simulate two and three

dimensional flows on multi-block structured grids by solving cell-centred integral forms of the

Reynolds-averaged Navier-Stokes (RANS) equations [10].

The Diagonalized Approximate Factorization (DAF) temporal advancement scheme was used

to achieve steady state solution convergence with a Courant-Friedrich-Lewy (CFL) regime

ranging from 0.1 to 3.0.

The fluid is treated as a thermally prefect mixture of gas that obeys empirically derived 3-

interval 9-coefficient caloric curve fits, for temperatures between 200 and 20,000 K.

The combustion of hydrogen was modelled using the 9 species and 18 finite rate reactions

described and validated by Drummond [11].

Turbulence was modelled using Wilcox’s k-ω model [12]. This model was shown to reliably

simulate mixing layers by Cutler et al. [13]. A second order accurate scheme was used for the

convective terms in the turbulence equations. Sutherland’s law is used to compute the

molecular viscosity. Thermal conductivity is computed from viscosity assuming constant

Prandtl number. The molecular diffusion coefficient is computed from viscosity assuming

constant Schmidt number.

Numerical modelling of the Razzaqi & Smart experiment is relatively straightforward due to

the simple geometry of the experimental model. The two-dimensional grids were separated

into two regions: the compression wedge and the combined rectangular combustor with

expansion surface. The process was performed using Pointwise version 16.03R2 [14].

Uniform inflow conditions, equivalent to the respective shock tunnel exit conditions, were

applied at the inflow boundary of the compression wedge region. The test section walls have

been modelled using the isothermal (300K) law of the wall boundary conditions (BC)

implemented in Vulcan (unless otherwise stated). Injected inflow conditions utilise the generic

subsonic inflow BC implemented in VULCAN. First order extrapolated BCs were applied at

the outflow boundaries.

Results

All simulations performed were based on the previously mentioned shock tunnel experiments

(refer to Table 1) except the nominal 5% oxygen enrichment case. This simulation used the

same injected flow conditions as the nominal 15% oxygen enrichment case but replaced the

75/25 (% by wt) O2/N2 mixture with air. A summary of the RANS simulations performed and

the corresponding Razzaqi and Smart experiments is detailed in Table 2.

Table 2: Summary of RANS simulations of the Razzaqi & Smart experiment

Simulation Shot Cells Turbulence

Model

Wall

Model

Chemistry

Model

Fuel-off, 2D coarse grid 9941 200K k-ω Yes No

Fuel-off, 2D fine grid 9941 686K k-ω Yes No

Fuel-off, 2D fine grid

without wall model 9941 686K k-ω No No

3D compression wedge,

coarse grid 9941 2.5M k-ω Yes No

Mixing-only 9952 233K k-ω Yes No

Fuel-on 9953 233K k-ω Yes 9×18 FR

5% Oxygen enrichment N/A 233 K k-ω Yes 9×18 FR

15% Oxygen enrichment 9944 233 K k-ω Yes 9×18 FR

Duct Entrance Flow Conditions

The initial stage of each simulation required determining the flow through the compression

wedge into the duct. The flow exiting this plane was used as the inflow conditions for the

second grid region (which models the duct and expansion surface).

Compressible two-dimensional flow (oblique shock and Prandtl-Meyer expansion) relations

were used to determine theoretical duct entrance conditions for each experimental shot for

comparison. A summary of the computed duct entrance conditions is detailed in Table 3.

Table 3: Summary of computed duct entrance conditions

Shot Method Temp.

(K)

Static Pres.

(kPa)

Velx

(m/sec)

ṁC

(kg/sec)

9941 Theoretical 1877 40.6 3200 0.791

9941 2D RANS 2060 40.8 3070 0.704

9952 Theoretical 1824 39.5 3160 0.782

9952 2D RANS 2000 39.9 3020 0.685

9953 Theoretical 1893 39.9 3330 0.781

9953 2D RANS 2130 42.0 3180 0.701

9944 Theoretical 1874 40.7 3200 0.792

9944 2D RANS 2060 40.8 3100 0.703

Injector Flow Conditions

Injector conditions were calculated using a quasi one-dimensional flow analysis assuming a

thermally and calorically perfect gas with the following boundary conditions: a total

temperature of 300 K, choked flow through the injector throat and a mass flow rate based on

experimentally measured values (refer to Table 1).

Validation

Simulations using a fine grid without a wall function, a fine grid with a wall function and a

medium grid with a wall function of the fuel-off case (Shot 9941) were performed to ensure

grid convergence. Comparison of the results obtained using these computational grids, based

upon pressure values along the duct centreline, showed acceptably minor variation, as shown

in Fig. 3.

Fig. 3: Air Only CFD and experimental duct static pressure comparison

A comparison between the experimentally measured and simulated static pressures along the

duct centreline was performed in order to validate the simulation, as shown in Fig. 3 for the air

only case (Shot 9941) and Fig. 4 for the combustion cases (Shot 9953 & 9944). The

magnitude of simulated and experimentally measured pressures is in close agreement.

However, the difference between simulated and experimental shock reflection locations

increases as the flow travels downstream towards the expansion surface. Three dimensional

effects may be responsible for these discrepancies, as discussed in a later section.

Fig. 4: Fuel into Air CFD and experimental duct static pressure comparison

The simulations predict the experimentally observed pressure rise along the duct for flows

which experience combustion, as shown in Fig. 4. Positioning of the shock reflections is more

difficult to discern from the experimentally measured pressures since fluctuations are not as

clearly evident for the combustion cases.

Recirculation Region

A noteworthy feature of the mean flow through the duct is the recirculation region which

forms behind the injection strut, shown in Fig. 5. This region contains low pressure

recirculated flow which causes injected flow to be under-expanded just downstream from the

point of injection. The injected flow expands until it meets the ingested flow travelling

through the duct. In the fuel-on case, a Mach disk then forms in the injected flow causing it to

become subsonic. In the nominal 15% oxygen enrichment case, the flow structure is altered:

the injected flow is recompressed by a series of oblique shocks rather than by a Mach disk.

The strength of the shear layer generated between the injected and ingested flows, which

should induce strong mixing between the streams, varies significantly between the cases.

Fig. 5: Axial velocity contours behind the injector strut

Mixing and Combustion Characteristics

Ignition, identified by the net production of OH- radicals, occurs approximately 40 mm

downstream from the point of fuel injection (both for fuel only and oxygen enrichment cases)

as can be seen in Fig. 6. The intensity of OH- production is much higher in the oxygen

enrichment case.

Completion of H2 combustion, identified by the net production (formation) of H2O, occurs

within 1mm downstream of initial liberation of OH- radicals (both for fuel only and oxygen

enrichment cases), shown in Fig. 7. Further, a significantly stronger reaction region occurs at

an axial displacement of 760 mm in the oxygen enrichment case.

Fig. 6: OH- production rate contours along the duct

Fig. 7: H2O production rate contours along the duct

The additional flow characteristics apparent in the oxygen enrichment case noticeably burn a

greater amount of the injected H2, shown in Fig. 8.

The combustion efficiency ( C ) of the combustor is give by the ratio of H2O mass flow

exiting the simulated combustor to the expected H2O mass flow produced by completely

combusting the injected H2:

2 2

2 2

,

,

H H O EXIT

C

H O H IN

M m

M m (4)

Where 2H OM is the molar mass of water and

2 ,H O EXITm is the mass flow of water exiting the

duct.

Fig. 8: H2 mass fraction contours along the duct

For the fuel only case, combustion can only occur between the injected H2 and the ingested

O2. However, oxygen enrichment cases provide a second source of O2 which is premixed with

the injected H2. This allows the injected H2 to combust with both the premixed O2 and the

ingested O2. For this reason, in the case of oxygen enrichment, combustion efficiency is not a

direct measure of the increased combustion strictly caused by the available H2 combusting

with the atmospheric O2.

It is reasonable to expect the stoichiometric amount of H2 premixed with the enrichment

oxygen will combust completely due to the high combustor temperature. Subsequently, this

amount of H2 cannot react with the ingested atmospheric O2 residing in the combustor. The

efficiency with which the remaining H2 (after premixed combustion) reacts with the

atmospheric O2 is defined as the compensated combustion efficiency ( ,C EPC ):

,

100%

100%

CC EPC

EP

EP

(5)

A summary of combustion parameters is detailed in Table 4.

Table 4: Summary of Combustion Parameters

Simulation ϕ

CFD

EP

CFD (%)

Temp.

(K)

ηC

Experiment

ηC

CFD

ηC,EPC

CFD

Fuel-on 0.84 N/A 2540 45 52.32 N/A

5% Oxygen enrichment 0.81 4.4 2610 N/A 57.98 56.04

15% Oxygen enrichment 0.81 14.1 2730 60 62.72 56.40

All three simulations have a similar equivalence ratio of almost unity. The simulation results

show for a premixed fuel percentage of 4.4%, an additional 5.7% of the injected H2 is burnt in

comparison to the fuel only (no premixing) case. This implies that an additional 1.3% of the

injected H2 has reacted with the ingested O2 due to the secondary effects of oxygen

enrichment.

As previously mentioned, the premixed fuel combustion reduces the amount of injected H2

available to mix and combust with the ingested O2. The compensated combustion efficiency

values show that both oxygen enrichment cases cause an additional 4% of the remaining

available H2 (after premixed combustion) to react with the ingested O2.

Three-Dimensional Effects

One possible explanation for the mismatch of pressure distributions between simulated and

experimental results could be the three-dimensional effects caused by the sidewalls of the

experimental model. To investigate the uniformity of the flow entering the duct in the cross-

stream direction, a three-dimensional simulation of the compression wedge was performed.

Cross-flow velocities of up to 8% of the axial flow velocities were observed in the results of

this simulation, shown in Fig. 9. This indicates that three dimensional effects may contribute

significantly to the discrepancies between the numerical and experimental pressure

distributions along the duct centreline. For example, changes in the axial Mach number within

the duct will alter the location and strength of the shock pattern.

Fig. 9: Static pressure through the wedge and cross-flow velocity entering the duct

Conclusion

A numerical investigation using Reynolds-Averaged Navier-Stokes simulations was

presented. This investigation provides new details on the flow physics within an oxygen

enriched scramjet combustor. The simulation results show that oxygen enrichment affects the

internal flow structure within the particular scramjet combustor. Specifically, oxygen

enrichment affects the mixing and combustion characteristics within the combustor. The

influence of oxygen enrichment on combustion is most significant in the ignition region.

Improvements to combustion efficiency can be made beyond the percentage of premixed and

injected O2. For example, an enrichment percentage of 4.4% produces an increase in

combustion efficiency of 5.7%.

References

1. Smart M.K. and Tetlow, M.R., “Orbital Delivery of Small Payloads using Hypersonic

Airbreathing Propulsion”, Journal of Spacecraft and Rockets, Vol. 46, No. 1, 2009, pp.

117-125

2. NASA “U.S. Standard Atmosphere 1976”, NASA-TM-X-74335

3. Heiser W.H. and Pratt, D.T., Hypersonic Airbreathing Propulsion, AIAA Education

Series, AIAA, Washington DC, 1994

4. Rudakov, A.S and Krjutchenko, V.V., “Additional Fuel Component Application for

Hydrogen Scramjet Boosting”, SAE Aerospace Atlantic Conference and Exposition,

Dayton, OH, USA, April 1990

5. Pike, J., “The Choice of Propellants: A Similarity Analysis of Scramjet Second Stages”,

Philosophical Transactions: Mathematical, Physical and Engineering Sciences, Vol. 357,

1999, pp. 2357-2378

6. Razzaqi, S.A. and Smart, M.K. “Shock Tunnel Experiments on Oxygen Enrichment in a

Hydrogen Fuelled Scramjet”, 16th AIAA/DLR/DGLR International Space Planes and

Hypersonic Systems and Technologies Conference, AIAA 2009-7429, Bremen, Germany,

Oct. 19-22, 2009

7. Stalker, R. J., Recent Developments with Piston Free Drivers, 17th International

Symposium on Shock Waves and Shock Tubes, 1990, pp 96-107

8. Razzaqi, S.A., private communication, Sep. 2010

9. White, J.A. and Morrison, J.H., “A Pseudo-Temporal Multi-Grid Relaxation Scheme for

Solving the Parabolized Navier-Stokes Equations”, AIAA Paper 99-3360, 14th AIAA

Computational Fluid Dynamics Conference, Jun. 28 – Jul. 1, 1999.

10. NASA, Vulcan version 6.1.0 User Manual, 2009

11. Drummond, J.P., “Numerical Simulation of a Supersonic Chemically Reacting Mixing

Layer”, NASA TM 4055, 1988

12. Wilcox, D.C., Turbulence Modeling for CFD, 3rd edn, DCW Industries, La Cañada, 2006

13. Cutler, A. D., Diskin, G. S., Drummond, J. P., and White, J. A., “Supersonic Coaxial Jet

Experiment for Computational Fluid Dynamics Code Validation” AIAA Journal, Vol. 44,

No. 3, 2006, pp. 585-592.

14. Pointwise Inc., Pointwise version 16.03R2 User Manual, 2010